1111

2 Pe bble m a t rix filt ra t ion in

3
4 Pa pua N e w Guine a
5
6
7 J.P. Rajapakse
8
9 Slow Sand Filters become inoperative during heavy rains
10
when the raw water supply becomes turbid. A novel pre-
1
treatment method called Pebble Matrix Filtration was
2 tested in the laboratory and then tried out in the field in
3 Papua New Guinea. Pre-treatment was shown to reduce
4 turbidity and faecal coliforms dramatically.
5

S
6 low Sand Filters (SSFs) are an during periods of heavy rain when the ward, first through a layer of pebbles
7 essential element of water treat- supply of raw water becomes turbid. A and then through a matrix of pebbles
8 ment works in many developing typical example is the Bumbu River in and sand mixture. The upper part of the
9 and developed countries. The World the Morobe Province, PNG, where pebbles-only bed has some pre-filtering
20111 3000 mg/l suspended solids is quite effect, but the reduction in suspension
Health Organization (WHO) recom-
1 common during rainy periods. The silts concentration occurs mainly in the
mends the use of slow sand filters for
2 and clays in muddy water clog the sur- lower part of pebble–sand mixed bed.
the treatment of drinking water, particu-
3 face layer of sand and fill the pores Initially it was accepted that the
larly for rural or village supply. The
4 between the sand grains, with a conse- ability of the PMF to cope with very
simple designs of slow sand filters per-
5 quent loss of flow capacity and a rapid high turbidity was just a useful empiri-
mit the use of local materials and skills
6 rise in head-loss across the filter. This cal phenomenon with no rational expla-
in their construction, operation and
7 shortens the filter runs, necessitating nation. However, the use of endoscopes
maintenance. No chemicals are used for
8 frequent cleaning, thereby disrupting in the laboratory revealed that the
their operation, and almost no instru-
9 the water supply. capacity of the PMF to accept high
mentation is required, so the cost of
30 It is therefore vital to protect the deposit loadings without a great drop in
imported materials and equipment can
1 SSF from such effects during monsoon pressure is attributable to flows over
be kept to a minimum. The SSF is
2 or heavy rainfall periods by the use of the pebble surfaces acting as a ‘wall-
very effective in pathogen removal and
3 pre-treatment methods. A novel pre- effect’, and to ‘lens-like’ cavities found
4 generally produces an effluent satisfy-
treatment method called Pebble Matrix underneath the pebbles. These
5 ing bacteriological drinking-water
Filtration (PMF), developed at Univer- additional flows (macro) are a signifi-
6 quality requirements. It therefore satis-
sity College London, proved to satisfy cant fraction relative to the flow pass-
7 fies the needs of a water-treatment these conditions and showed promise in ing through the sand (micro) and main-
8 facility in low-income countries. the laboratory.3 Other books highlight tain a high effective permeability
9 Handbooks containing guidelines for the developments in slow sand filtration throughout the bed, allowing an incom-
40111 the design and construction of these from around the world.4 ing suspension to penetrate to the
1 simple but efficient natural filters have After the laboratory trials in London deeper, as yet unclogged, sand. The
2 been published by WHO and the Inter- using kaolin clay suspensions in Lon- PMF was to perform as a pre-treatment
3 national Reference Centre for Water don tap water, the PMF had to be unit before SSF, therefore it was not
4 and Sanitation (IRC).1,2 tested out using natural raw water in expected on its own to produce potable
5 There are no SSFs operated by the PNG. It was also necessary to examine water. Hence, filter cleaning was
6 water board or other water companies the feasibility of construction, operation carried out by ‘two drainage cycles’
7 in Papua New Guinea at present. This and maintenance of a PMF under rural (first by draining down the supernatant
8 may be because the simple-technology PNG conditions and then to assess its water in the filter and then refilling
9 image has resulted in it being con- suitability as a pre-filter for SSF in with raw water while the drain valve
50 sidered generally inappropriate by treating surface waters for village water was closed, and then reopening the
1 some designers, but such perceptions supply. The design and construction of drain valve) followed by a backwash
2 are rapidly changing in the light of the system were carried out by students with raw water under gravity.5
3 the rising demand for higher drinking of the Department of Civil Engineering
4 water quality standards. In the past at University of Technology. First PNG trial plant
5 10–15 years, there has been a renais-
6 sance of interest in the potential use of In order to test the technology in the
Pebble Matrix Filter (PMF)
7 slow sand filtration throughout the field, a PMF unit and two SSF units
8 world. The PMF can be described as a crude were constructed at Ambuasuz village,
9 A long-standing problem with slow two-layer filter, where a turbid suspen- 14 miles outside Lae. The PMF pre-
60111 sand filters is that they deteriorate sion approaching the filter flows down- treated water was fed into one SSF and

12 Vol. 21 No. 4 April 2002

 household water security
1111 the other SSF acted as a reference fil-
2 ter. About 2.5 m of head was allowed
3 between the dam outlet and the PMF
4 for gravity flow and the same head was
5 used for backwashing the PMF.
6 The two SSF tanks were of 0.9 m
7 diameter, 2.5 m high and were assem-
8 bled on site, by joining half-culvert
9 rings made of corrugated galvanized
10 iron. Sand of ‘effective size’ d10 = 0.3
1 mm was selected from a local beach as
2 the filter medium (here d10-= Hazen’s
3 effective grain size in mm, relative to
4 which 10 per cent of the sample is
5 finer). Transparent plastic tubes of
6 5 mm diameter were fitted at different
7 heights for head-loss monitoring
8 through the bed. The filters were oper-
9 ated at a filtration rate of 0.2 m/h.
20111 One PMF tank of 0.6 m diameter
1 and 1.4 m tall made of half-culvert Figure 1 Details of the PMF column, Lae plant
2 rings of corrugated galvanized iron
3 was assembled on site by the same Suspended solids and bacteriological fifth month. Due to very low head-
4 method used in constructing SSF analysis were carried out at the losses the PMF was operated for over
5 tanks. Rounded smooth pebbles of Department’s Public Health Laboratory two months without backwashing, con-
6 approximately 50 mm diameter were with occasional samples sent to the sequently the filter pores became
7 handpicked from a local river aggregate National Analysis Laboratory (NAL) clogged. The bed required manual
8 site. For the preliminary experiments, a for confirmation. cleaning due to mud-ball formation
9 pebble bed depth of 80 cm was partly after six months of operation. These
30 infilled to a depth of 40 cm with Results and discussion experiments proved that with fairly
1 medium pool-sand of grade 16/30 low-turbidity raw waters, the combina-
2 (0.50–1 mm), purchased from a local The plant was commissioned success- tion of a PMF and a SSF would allow
3 supplier. Similar arrangements to the fully in late 1999. Unlike in the labora- up to six months of operation without
4 SSF using manometers were made for tory trials, large amounts of algal cleaning if desired, a condition that
5 the head-loss monitoring through the growths were noticed both in the PMF may well suit rural PNG conditions.
6 PMF bed. The assembly took less than and in the reference SSF, whereas the Although the field backwash rate
7 half a day for two people to complete. SSF fed from the PMF filter was virtu- obtained with the available head was
8 The details of the PMF unit are shown ally free from growths. As a result, the much lower (10 m/h) compared to labo-
9 in Figure 1. head-loss in the former SSF was as ratory backwash rate of 50 m/h, a satis-
40111 The laboratory under-drain system high as 140 cm compared to 10 cm in factory cleaning was achieved with two
1 comprised a cone-shaped bottom filled the latter after four months. At the end drainage cycles followed by backwash-
2 with marbles for uniform distribution of of the fourth month the reference SSF ing once for five minutes.
3 backwash flow and ease of construc- was cleaned, and the sand washed and Since the construction of the raw
4 tion, but in prototype filters several replaced, whereas the SSF that received water storage reservoir the inlet tur-
5 types of under-drain systems were con- PMF pre-treated water continued in bidity to the PMF rarely exceeded
6 sidered: perforated pipes, corrugated operation without cleaning, resulting in 50 NTU (Nephelometric Turbidity
7 pipes, etc. The perforated pipe system a final head-loss of only 15 cm by the Units), whereas turbidities of
8 was chosen for the under-drains in the
9 field test (see Figure 2). The PMF tank
50 was taken to the site and filled
1 manually with pebbles, infilling the
2 spaces between pebbles with sand,
3 layer by layer about 10 cm thick at a
4 time to the required depth.
5 The raw water and filtrate turbidities
6 were monitored using a portable
7 turbidimeter, model ‘Hach-2100P’.
8 Head-loss measurements through filter
9 beds were carried out manually along
60111 the manometers using a tape measure. Figure 2 PMF underdrains of Lae and UCL Žlters

Vol. 21 No. 4 April 2002 13

 household water security
1111 100–200 NTU were very common struction and commissioning of a field- tal work. Markham Culverts Ltd., Lae,
2 prior to construction. Filtrate turbidities scale plant in Lae confirmed the donated all culvert rings for filter
3 of both SSFs were below 0.3 NTU. feasibility of constructing such a system construction. The help of the Ambuasuz
4 Faecal coliforms of 1300–1600 CFU under PNG conditions and eliminated village community, staff of the Civil
5 (Colony Forming Units) and total some doubts regarding cleaning diffi- Engineering Workshop, the Department
6 coliforms of 2200–2800 CFU in the raw culties of the PMF. Preliminary field and the Appropriate Technology and
7 water were reduced to zero faecal and tests also confirmed the laboratory scale Community Development Institute
8 total CFU in 100 ml samples of both experiments and all the indicators are (ATCDI) is very much appreciated.
9 SSFs. These preliminary results show that the PMF is a useful tool in protect-
10 that the PMF is a useful tool in prolong- ing an SSF in all weather conditions,
References
1 ing the life of an SSF and the system especially during rainy periods. It is
2 can be used as a package in rural drink- desirable to incorporate a larger-scale 1 Smet, J.E.M. and Visscher, J.T. (ed.),
3 ing water treatment plants. Some inter- treatment plant into a future treatment (1989), Pre-treatment methods for com-
4 mittent simulated high turbidity experi- facility and the Water Board is very munity water supply, WHO-IRC, The
5 ments (made by adding silt to the raw interested in the developments so far. Hague.
6 water) were also conducted with the 2 Van Dijk, J.C. and Oomen, J.H.C.M.
7 PMF. With the support of ‘Eda Ranu’, a About t he a ut hor (1978) Slow sand filtration for com-
munity water supply in developing coun-
8 PNG water company, further investiga- Dr J.P. Rajapakse is a lecturer in the Department of
tries; a design and construction manual,
9 tions using naturally high-turbidity Civil Engineering, University of Technology, Lae, Papua
Technical Paper No.11, WHO, Geneva.
20111 water of the Bumbu River are being New Guinea.
3 Rajapakse, J.P. and Ives, K.J. (1989)
1 conducted at a new site near Butibum
‘Pebble Matrix Filtration’, in Smet and
2 village, a short distance from Lae City. Acknowledgements Visscher op. cit.
3 4 Graham, N.J.D. and Collins, R. (ed.),
4 The original laboratory experiments were
Conclusions (1996), Advances in slow sand and alter-
5 financed by University College London native biological filtration, John Wiley
6 A non-chemical, simple filtration and DFID. Design and construction were and Sons, Chichester, UK.
7 system comprising an SSF with a PMF carried out by final-year undergraduates 5 Rajapakse, J.P. and Ives, K.J. (1990)
8 as pre-filter proved to satisfy the water of the Department of Civil Engineering, ‘Prefiltration of very highly turbid
9 treatment requirements of rural con- University of Technology. Billy Ben- waters using pebble matrix filtration’,
30 ditions in PNG. The satisfactory con- jamin carried out most of the experimen- Journal of IWEM, 4 (2): 140–7.
1
2
3
4
5
6
7
8
9
40111
1
2
3
4
5
6
7
8
9
50
1
2
3
4
5
6
7
8
9
60111

14 Vol. 21 No. 4 April 2002